Following are the areas of research that Compiler Research Group is focused on:
Automatic Differentiation (AD) is a useful technique in scientific research fields like machine learning and computational physics. AD enables the automatic computation of derivatives of functions with high precision and efficiency. A notable implementation of AD is the Clad plugin for the Clang compiler. This integration not only simplifies the process of differentiation but also enhances the performance and accuracy of numerical computations in scientific applications.
In scientific research, where intricate mathematical models are prevalent, the utilization of AD through tools like the Clad brings a new level of sophistication and speed to derivative calculations. By leveraging AD within C++ compilers, researchers can focus more on the scientific aspects of their work rather than getting bogged down in manual differentiation tasks. This automation not only accelerates the development process but also ensures that computations are error-free and consistent.
Compiler as a Service (CaaS) is an evolving technology that redefines the traditional approach to compilers by providing a service-oriented architecture. Instead of treating the compiler as a black box, the CaaS approach helps open up the functionality to make it available as APIs. This gives developers unprecedented control and insights into the compilation process, while being able to use lightweight APIs for simpler workflows and diagnostics, helping create sophisticated applications more efficiently.
Practical applications of CaaS include deeper and interactive program analysis and conversion from one programming language to another (e.g., C++ and Python).
Despite its high performance capabilities, C++ is not the first programming language that comes to mind for rapidly developing robust applications, mainly due to the long edit-compile-run cycles.
Ongoing research in projects such as Cling, Clang-REPL, etc. aims to provide practically usable interactive capabilities to the C++ programming language. The goal is to enable dynamic interoperability, rapid prototyping, and exploratory programming, which are essential for data science and other scientific applications.
Following are some practical applications of a “C++ Interpreter,” so to speak:
In Data Science: Interactive probing of data and interfaces, making complex libraries and data more accessible to users.
In CUDA: The Cling CUDA extension brings the workflows of Interactive C++ to GPUs without losing performance and compatibility to existing software.
In Exploratory Programming: rapid reproduction of results, which is crucial during the exploratory phase of a project.
In Jupyter Notebooks: Interactive C++ can be integrated with Jupyter Notebooks, providing a swift prototyping and learning experience for C++ users.
Language interoperability helps programmers get the best of both worlds, with the ability to work with a high-performance language (e.g., C++), and at the same time, take advantage of a more interactive one (e.g., Python), while helping them identify each other’s entities (like variables and classes) for seamless integration.
This interoperability can be achieved by libraries like CppInterOp, which expose APIs from compilers like Clang in a backward-compatible manner. By enabling interactive C++ usage through the Compiler-As-A-Service, CppInterOp simplifies complex tasks such as “language interoperability on the fly”.
The practical implications of language interoperability include the growing need for systems in data science to be able to interoperate with C++ codebases. By providing automatic creation of bindings on demand, tools like CppInterOp enable Python to interoperate with C++ code dynamically, instantiate templates, and execute them efficiently. This dynamic approach not only improves performance but also simplifies code development and debugging processes, offering a more efficient alternative to static binding methods.
We are part of the
Princeton Department of Physics
and
Princeton Institute for Computational Science and Engineering
(PICSciE).
Contact: David Lange; Vassil Vassilev
This project is supported by the National Science Foundation under Grant OAC-1931408 (2020-2023) and OAC-2311471(2023-2026). Any opinions, findings, conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.
© 2019 Princeton University. Site made with Jekyll; Credit to Allan lab for this site template.
